Authors

H. Karacali, E. Cebel, TTTech Auto, Izmir, Turkey

Year

2026

Print Info

Production/Publication ÖVK

Summary

The growing complexity of SDVs, driven by higher autonomy levels and massive sensor-fusion data streams, makes reliable anomaly detection essential for maintaining safety and system integrity in safety-critical domains. Anomaly detection in traditional SDVs, when executed within central processing units, results in performance and reliability bottlenecks. This study introduces a holistic and progressive concept from TTTech Auto that advances the 4SDV (Systems, Safety, Security, and Software-Defined Vehicle) paradigm through the integration of Edge AI into zonal architectures. Zonal controllers detect anomalies using NPU-based Edge AI, which are deterministically validated by MCUs before forwarding only confirmed cases to the central compute. This dual-layer design combines AI adaptability with deterministic, rule-based verification to ensure fail-operational reliability in the event of faults. Verified anomalies are disseminated via DDS/Zenoh under QoS guarantees and supported by Time-Sensitive Networking (TSN), optimizing bandwidth and ensuring deterministic low-latency response. The approach mitigates central compute overload, embeds system-wide safety and security from the outset, and ensures compliance with functional safety (ISO 26262) and cybersecurity (ISO/SAE 21434) standards. In addition, it introduces a certifiable zonal intelligence layer into the 4SDV framework. Edge AI integrated 4SDV brings a smarter, safer, and more reliable perspective to the modern SDV paradigm.

ISBN

978-3-9504969-5-6

DOI

https://doi.org/10.62626/98j9-sfz6 Copy link

Order Members Login